uh, sample margin of errors is a straightforward calculation. The MoE is based on sample size and population, nothing more.
The Pew poll had a MoE o 2.2, not 5%. Don't know where you saw 5%
http://www.people-press.org/files/legac ... elease.pdfThe so-called problem with phone sampling you demonstrate doesn't seem to be evidenced strongly. What we do seem to know is that live calls and cell phone polls tend to lean more democratic (and probably be more accurate) and that there's potential for some loss of Hispanic vote due to non-english speakers. But these numbers aren't large, either.